EECS225B Problem Set 5

Emaad Khwaja

Problem 1

The Adaptive Median Filter implementation is shown below.

Below are the results of the filtering methods. For the Wiener method, a size of 15 was used with a noise power of 300. The adaptive median filter used Sxy = 3 and Smax = 9.

I decided I liked the smoothign effect seen on the car of the Wiener filtered image compared to the others, but I did not like the level of detail lost in the sand in the sky. This was preserved in the adaptive median filter. To create a nicer hybrid, I took a gaussian lowpass of the noisy image and a gaussian highpass of the adaptive median image and merged them together. ResultA is the most aesthetically pleasing image in my opinion.

Problem 2

The iterative convolution method is shown below.

Below are the results of applying the iterative deconvolution method and a Wiener filter for noise removal. I found that it was better to perform the noise removal step after the deconvolution, as reversing this order often left too much information lost for the deconvolution to be effective.
I think ResultC has the best compromise between ringing and sharpness.